According to the atmospheric electric intensity data of the first happen CG flash(shortened as "thunderstorms") and the first unhappen CG flash(shortened as "un-thunderstorms"), the FFT(Fast Fourier Transform) method has been adopted to converse time data to frequency domain, which is used to analyze the amplitude spectrum of the data. The results show that there is an obvious spectrum amplitude distribution difference between the two sets of data, one of which belongs to 30 minutes before the first flash in thunderstorms and the other is related to arbitrary 30 minutes in un-thunderstorms. Furthermore, the two kinds of weather spectrum amplitude distribution may have followed some certain rules. Therefore, on the basis of sample classification, this paper regards two types of averaged weather spectrum amplitude as classification index; uses Euclidean distance method on 63 unknown weather samples to classify the data and examine the results, which shows that this new method can be considered as a reference for the approaching and short-term forecast of the first CG flash.